EXT3 Aborting Journal / Kernel Soft Lockup Help
I have a system running CentOS 5.5 (194.32 Kernel) that has been working fine for about 8 months. I found it stuck one morning dumping a Soft Lockup Kernel message every 10 seconds and have been trying to find out what is going on - here is what I've done and found.
System is a Core i3 with 4Gb DDR3 ram and 2 1.5Tb SATA drives - everything is less then 1 year old with not too much load (this is a backup system for my server - a twin and the other works fine) Intel Mobo: DM55HC Rebooting the system (after doing a fsck on the drives with lots of errors) it runs for about 20 hours or so and then drops back into this problem. Prior to going into the soft lockup it reports that the Journal has aborted on one drive or the other and drops into a Read Only FS. There is nothing in the logs apart from the aborting journal message and the soft-lockup) Tried fsck with badblocks check on the 2 mountpoints for both drives (dunno how to check the swap partition) - nothing I could see there. I have tried memtest for about 36hours. No errors I ran smartctl long tests on both drives. No Errors I'm beginning to think it's the SATA controller on the motherboard but I have no idea how to test this. I'd appreciate any thoughts of what I can do next. |
It's the backup; can it come offline? What's the output of smartctl?
BTW have you tried remounting ext2? |
Yes, in fact, it's offline now - I have tried rebooting and remounting and running at different run levels several times - When I switched to "init 1" and unmounted everything it seemed to sit there without incident for 24h - which would make me think it's not a CPU thing but as soon as the drives come into play it seems to be just a matter of time before it fails.
Ambient, motherboard, and CPU temps are all < 30C so I don't think it's overheating either. The output from smartctl was basically that the test had completed and there were no errors detected - that's all it said. I can run a specific smartctl report if that would help just let me know. |
Ok, basics all good.
Next I would look into the specific error. Exactly what has gone wrong when a kernel goes into soft lockup? I've never seen that. Check the kernel bugzilla. Can you make the journal external? e.g. the journal for drive a on drive b, & vice versa? tune2fs -J I think. You might end up emailing support - in the drive manufacturer, chipset or CentOS. I would also change what you can painlessly change for the sake of changing it, so you can eliminate it from suspicion. |
After asking around a lot I may have a solution - someone suggested turning off acpi in the kernel they though it might be an IRQ conflict and I had seen some messages about a certain IRQ getting disabled upon boot. I have tried this and I'm waiting to see what the system does in 24-36 hours.
I'll post a followup in either case. |
Quote:
|
More info...
Tried the IRQ and Kernel options to no avail... Since then I have also tried the following: Installing a new Centos5.5 on a different HD (same make/model) and let the system run for 3 days - no problems at all with this. Tried booting again from the old HD and got the same problems after various lengths of time. I have just found something new though - after the system crashes and I get into the repair filesystem prompt and start running fsck I can do this over and over and every single time I run it I always get errors being fixed - I have run repeated SMART tests and badblocks tests all showing that everything is fine - but every time I run fsck it finds and corrects errors. This makes no sense to me. Any ideas? Detailed Hardware description: Intel DH55HC Motherboard Core i3 530 @ 2.93GHz 4Gb OCZ DDR3 - 1333 Ram 1Tb ST31000528AS Segate HD (boot Drive) 1.5Tb WDC WD15EARS-00Z5B1 WD HD (data Drive) I will try to get exact error messages but it's hard because the system keeps failing. |
Quote:
|
Quote:
Quote:
Quote:
|
I have been running long self tests in smartctl I finally managed to capture some results.
I'd love to replace the drive - but to do that I must get segate's seatools to give me a code so I can get an RMA or somehow convince them it IS the drive failing - unfortunately when I use their seatools bootable ISO the thing crashes (which seems to be a common problem) and so looking for an alternative the SMART output below really isn't helping my case (the results here look just fine to me no?). du -sh /lost+found gives 172K but doesn't seem to be increasing. Code:
smartctl version 5.38 [x86_64-redhat-linux-gnu] Copyright (C) 2002-8 Bruce Allen and the other drive... Code:
smartctl version 5.38 [x86_64-redhat-linux-gnu] Copyright (C) 2002-8 Bruce Allen |
Update: Just found a firmware update for the ST31000528AS - I have applied it and am running another smartctl -t long
Strange because if you search for the update on the segate site using the drive serial you don't find it - but doing a search on google you do - else I would have found this ages ago. ... again I'll post an update when I have one. |
now the other drive goes into error mode again - this is from the WD - seems highly unlikely to me that 2 drives from 2 different manufacturers would fail at exactly the same time both showing no SMART errors nor badblocks errors. This must be the Mobo or the OS no?
Code:
Feb 8 06:03:00 backup kernel: EXT3-fs error (device sdb2): ext3_lookup: unlinked inode 281804899 in dir #298582113 |
More information.
I tried re-installing with CentOS 5.5 on new hard drives - as soon as I started the backup running - (takes about 4 hours) the same problem cropped up - this confirms for me without any doubt that it's NOT the hard drives that are the problem. I believe it's either the OS or the Motherboard - Intel won't support the Mobo with Linux so I have now installed Windows 7 on the system and am running the backup that way to see if any problem crops up. Not really asking any question at this point (unless someone has an idea) just updating the thread in case anyone else has a similar problem ever. |
Thanks, definitely interesting. Could help to post motherboard detailed version and HW specs, especially used chipset nfo (or update the http://www.linuxquestions.org/hcl/).
|
I've had similar on 2 occasions
1. A mismatch of drives (Seagate & anything else) on the same ide lead produced 'ringing' so things saw logic levels that weren't there, and the fs threw errors. 2. logic levels on a bus drifting out of spec, at which point the device stopped working. |
All times are GMT -5. The time now is 08:35 PM. |